# Replay runs in bulk

> Recover many affected runs after an incident without restarting them one by one.

After fixing an incident, replay the affected runs by function, time range, and status. Replay creates new executions, so check which charges, messages, or records could repeat. Follow [Idempotency](/docs-markdown/durable-execution/guides-and-advanced/idempotency) when a repeated charge, message, or record update would cause harm.

1. Deploy and verify the fix.
2. Open the function in the dashboard and choose **Replay** from **All actions**.
3. Name the replay after the incident. Select its time range and the statuses to include.
4. Review the selection. Include successful runs only if they also need recovery.
5. Start the replay and watch its progress.
   Replay spreads the selected work over time. Use a narrow time window and confirm the affected functions before starting. For one run, use [Rerun a run](/docs-markdown/platform-and-operations/rerun-a-run) instead. For events skipped while a function was paused, select the **Skipped** status after resuming that function. Historical availability follows your plan's event retention; check pricing rather than assuming an unlimited replay window.